investment climate and fdi in developing countries: firm ...economics.ca/2008/papers/0339.pdf ·...
TRANSCRIPT
Investment Climate and FDI in Developing Countries: Firm-Level Evidence
Tidiane KINDA*
CERDI-CNRS, Université d’Auvergne
This draft (February 2008)
Abstract
This paper shows that investment climate constraints jeopardize Foreign Direct
Investment (FDI) using firm level data for 77 developing countries. Investment climate
constraints are characterized firstly by physical and financial infrastructure problems and secondly
by including human capital constraints and institutional problems. The main results are robust to
alternative definition of FDI, introduction of additional explanatory variables, and some
breakdown analyses (exporter versus non exporter firms, different sectors of activity and
comparison between Sub-Saharan Africa and other developing countries).
Keywords: FDI, Investment climate, Firm level data, Developing countries.
E-mail : [email protected] *The author is grateful to Leandre Bassole, Jean Louis Combes, Alain de Janvry, Elisabeth Sadoulet, Luc Desire Omgba and participants at the Joint Doctoral Seminar organized by CERDI, Université d’Auvergne and CES Université de Paris I-Sorbonne.
2
Introduction
Although the last two decades have been marked by a surge of private capital flows to
developing countries, developed countries remain the first destination for Foreign Direct
Investment (FDI). In the developing world, some countries succeed in attracting foreign capital
while others are more marginalized. Even if there is some evidence of its negative effects, FDI is
recognized to have beneficial effects on local firms and the economy at large. For instance, FDI
gives more resources, facilitates technological and managerial knowledge transfers to the host
countries, develops their international import and export network, creates new job opportunities,
and promotes economic growth. Why are some countries more attractive to FDI than others?
This is an important issue in economics, business and politics for which it is important to analyze
and understand the forces driving FDI location.
A complete and often used conceptualization of FDI determinants is the Eclectic
paradigm (Dunning, 1980, 1993). This paradigm provides a framework that groups micro and
macro level determinants in order to analyze why and where Multinational Enterprises (MNEs)
invest abroad. This framework is based on Ownership, Location and Internalization advantages
known as OLI. Our study focus on the location aspect of OLI framework according to which
MNEs invest in a foreign country in order to get advantages based on location (lower factor cost,
lower trade cost, etc.). There are three major reasons why a firm invests abroad following the
location advantage of the OLI framework. The first one is to exploit and export natural resources
and resource-based products. The motivation of these resource-based investments is mainly
resource availability. The second reason is to supply the domestic market of the recipient country
through an affiliate: Horizontal FDI (HFDI). In this case of market-oriented investment, gains in
trade costs and strategic advantages (intangible assets) should be important compared to the cost
of setting up a new plant. The third reason of FDI is to delocalize all or a portion of the
production process (production of components, and increasingly service activities such as call
centers) in order to benefit from low cost: Vertical FDI (VFDI). This kind of FDI probably
3
occurs when firms can break down their production process into different parts and different
locations according to factor costs in these locations. The determinants of vertical and horizontal
FDI thus differ, and the effects of the same variables can also be different according to the type
of FDI. A typical example is trade cost which positively affects HFDI and negatively VFDI.
Since it is very difficult to divide data into VFDI and HFDI, most studies use the aggregate FDI
with empirical analyses giving some average effects. Our study follows this pattern firstly and
then distinguishes the two types of FDI.
The question of what factors determine the location of FDI has received a much
attention in the literature. Most of the studies on FDI location using micro-level data focused on
variables such as R&D, factor cost differences, advertising expenditures, wages, trade cost,
market size or taxation (Disdier and Mayer 2004, Carr et al. 2001, Yeaple 2003, Hanson et al.
2001). These predictors are intuitive since these studies have focused on developed countries
(except China). For example, the availability of cheap labor or a large local market may be
important factors attracting foreign investment. However with the presence of deficient
infrastructure (road, electricity and telecommunication), high financing constraints, weak
institutions or lack of skilled labor as in some developing countries, these countries may not
necessarily exhibit the most attractive investment environment.
The role good institutions play in attracting FDI to developing countries is crucial. The
probability that foreign investors get return of their investments is fundamental in their decision
to invest in a country or not. Secure property rights, political stability and lack of corruption
allow markets to properly function, therefore attracting MNEs (Urata and Kawai 2000, Disdier
and Mayer 2004, Du et al. 2007). It is generally believed that the availability of skilled workers
positively affects developing countries’ attractiveness to foreign capital. In reality, according to
the type of FDI (Vertical FDI or Horizontal FDI), MNEs look for unskilled cheap labor or
skilled more expensive labor force. Yeaple (2003) finds that U.S. MNEs that invest in skilled
labor-abundant countries are skill-intensive industries while countries with a low-skilled labor
4
force receive more non-skilled intensive MNEs. Urata and Kawai (2000) and Fung et al. (2002)
get similar results.
Most of FDI location studies using firm level data in developing countries have focused
on China. This paper introduces firm level data for a large sample of developing countries in
order to assess the determinants of FDI with a focus on infrastructure, institution and human
capital. Contrary to other studies (except studies on US affiliates), this paper considers foreign
affiliates located in developing countries. The main results show that physical infrastructure
problems, financing constraints, and institutional problems discourage FDI to developing
countries, and especially to Sub-Saharan African countries.
The remainder of the paper is structured as follows. The first part reviews the theoretical
relationship between physical infrastructure, financial development and FDI location. The second
part presents descriptive and statistical analyses. The third part is devoted to econometric
analyses and results. The last part offers conclusions.
I. Physical Infrastructure, Financial Development and FDI
Well-developed infrastructure is critical to attract capital and promote economic growth.
Infrastructure availability is one of the key elements needed to run efficient business. In
manufacturing or services, good provision of infrastructure reduces transaction costs by allowing
entrepreneurs to easily connect with their suppliers and customers. A great number of studies in
developing countries have shown the importance of infrastructure for private capital
attractiveness using aggregated country level data (Asiedu, 2002; Ngowi, 2001). Most of the
studies that use disaggregated data to explore how infrastructure affects FDI location in
developing countries have focused on China and its provinces. (Cheng and Kwan 2000, Sun et al.
2002, World Bank 2006). A small number of studies include developing countries other than
China. Urata and Kawai (2000), based on Japanese Small and Medium Enterprises (SMEs)
location find that infrastructure is particularly important for developing countries FDI
5
attractiveness. Deichmann et al. (2003) in a study of 293 foreign firms location in Turkey find
that infrastructure development increases the probability of MNE location.
Do foreign firms need locally developed financial services? This question has not been
paid particular attention in the literature compared to other determinants of FDI location (wages,
market size, etc.). The importance of financial services for foreign firms is double. Like local
firms, foreign firms can use financial services for overdraft facilities, loans or payment to their
suppliers of intermediate goods. A developed financial market also facilitates financial
transactions between foreign firms and their customers and employees in the host country. In
general, financial development is also an engine of economic growth, providing better business
opportunities for customers and local and foreign firms. Few studies have linked FDI location to
financial development in developing countries and these studies in general find that financial
development encourages FDI (Jensink and Thomas 2002, Deichmann et al. 2003).
Dollar et al. (2006) analyze the importance of investment climate on export and FDI
probability for eight Latin American and Asian countries using firm level data. Their conclusions
are drawn for all investment climate variables together, which include physical and financial
infrastructure variables without giving specific effect of a particular variable. The authors
conclude that better investment climate in general encourages FDI.
II. Data and descriptive statistics
The data are drawn from enterprise surveys1 in developing countries conducted by the
World Bank. The surveys’ year ranges from 2000 to 2006 depending on the country. This analysis
includes 77 developing countries and 33,604 firms including 4,660 foreign firms. The dependent
variable (FDI) takes the value 1 if at least 10% of firm’s capital is foreign (following the IMF
standard of FDI definition) and 0 otherwise. The explanatory variables of first interest (physical
1 The enterprise surveys are the Investment Climate Analysis (ICA) and the Business Environment and Enterprise
Productivity Survey (BEEPS). For each country, the survey has been carried out between 2000 and 2006.
6
and financial infrastructure) include firm’s judgement of transport, electricity, and access to
finance problems. They also account for financial development through the share of informal
sources of financing (money lender, family and friends) in firms’ working capital (accounts
receivable, inventories and cash). As physical infrastructure variables, we also retain firms’ use of
e-mail and internet in their interactions with clients and suppliers giving an idea of firms’ access
to telecommunication. In accordance with the theory, some control variables (institutional
problems; lack of skilled workers; agglomeration, age and size effects) are included2. Appendix 1
gives the name and definition of all variables.
The proportion of foreign investors varies according to the sector3 (figure 1). High values
sectors such as electronics, metal and machinery, chemicals and pharmaceutics may attract more
foreign investors compared to other sectors in developing countries. Beyond the expected profits
in these sectors, they require a large amount of investment during set-up and operation. In this
sample, 43% of firms in the electronics sector and 19% in the chemicals and pharmaceutics
sectors are partly or fully-owned by foreign investors. Other sectors have on average 13% of
foreign firms except the leather sector in which firms are 96% locally owned.
2 Explanatory variables have been chosen for their economic relevance but also according to their number of non
missing values.
3 The sectors include in this analysis are textile, leather, garment, agroindustry (including food and beverage), metal &
machinery (including automobile), electronics, chemicals & pharmaceutics, wood & furniture (including paper), non-
metallic & plastic materials, retails & wholesale trade and services (excluding retail).
7
Figure 1: Share (%) of Foreign Firms by Sector
0
10
20
30
40
50
Leat
her
Wood
and
Fur
niture
Garm
ents
Textile
Servic
es (r
etail
exc
lude
d)
Agroi
ndus
try
Non-m
etall
ic & p
lastic
mat
.
Retail
and
who
lesale
trade
Met
al and
Mac
hiner
y
Chem
icals
and P
harm
aceu
tics
Electro
nics
Foreign and local firms in developing countries face many constraints when investing,
operating or expanding a business. According to firms perceptions in the enterprise surveys,
financing is ranked as the most important investment climate constraint for firms (local and
foreign), and foreign firms locate more where financing constraints are low (figure 2 and 3).
Figure 2: Ranking of Investment Climate Constraints
(all firms)
Figure 3: Foreign Ownership and Financing Constraint
0 0,5 1 1,5 2
Finance
Tax
Corruption
Crime and disorder
Skills of w orkers
Electricity
Trade regulation
Labor regulations
0 0,05 0,1 0,15 0,2
No obstacle
Minor obstacle
Moderateobstacle
Major obstacle
Very SevereObstacle
8
The interest variables (physical infrastructure constraints approximated by
telecommunication problems and financing constraints approximated by the use of informal
finance) are more objective variables compared to enterprises perception of constraints used in
figure 2 and 3. Using these “more objective” variables, the following graphs (figure 4 and 5)
highlight the fact that foreign firms (FDI) locate less where telecommunication problems are
higher and also where firms rely more on informal source of financing (friends, family) for their
business.
Figure 4: Telecom Problems and FDI Figure 5 : Informal Finance and FDI
0,0
0,2
0,4
0,6
No Yes
FDI
Tel
ecom
pro
blem
s
0,0
2,0
4,0
6,0
No Yes
FDI
Info
rmal
fina
nce
A deeper analysis based on econometric estimations and including control variables is
needed to go beyond these basic findings.
III. Econometric analysis
III.1. Estimation
Contrary to classic FDI location studies which consider the different possible location of
affiliates for each MNE, our study takes the side of all affiliates in a country. In former studies,
the dependent variable takes the value of 1 if the MNE chose the country as location of its
affiliate and 0 for other alternative locations. Most of the time, alternative locations are restricted
to a group of countries because of data availability or aim of the study. This implies exclusion of
some alternative locations which can lead to bias. In this study, we take the entire sample of firms
in one country and compute the probability for each firm to be foreign, given the characteristics
9
of different regions in the country. We expect that countries, and within a country, regions with a
better investment climate attract relatively more foreign firms given their better profit potential.
The equation to be estimated can be written as:
1 2 3 4ijk ijk ijk i j ijkFDI X Z V Uβ β β β ε= + + + +
FDIijk indicated if firm k in country i and sector j is foreign owned or local. Xijk is a
matrix including structural problems (physical and financial infrastructure problems, lack of
human capital and low quality of institutions). Zijk is a matrix of other determinants of firm
location (agglomeration effects, taxes, trade regulations or firms’ specific factors such as size or
age). By including Vi and Uj that are respectively country and sector fixed-effects, we explain the
regional variation.
The explanatory variables of the probability that a given firm is foreign or local are firm
level information or firms’ perception of investment climate constraints. More productive or
efficient firms (in general, foreign firms) can have a smaller feeling of investment climate
constraints compared to less productive firms (in general, local firms). The “same” investment
climate can then be assessed differently according to firms’ performances and resources. The
perception data can also be sources of measurement errors. These can lead to endogeneity, so we
define instruments that are the sector-region averages for each endogenous variable4. We also
consider the sector-region average of the fact that the firms’ annual financial statements are
reviewed or not by an external auditor as an instrument for the financial infrastructure variable.
Enterprise surveys include many variables explaining the same phenomena. One example
are variables related to financing constraints that include access to financing as collateral
requirement and, access to financing as the share of firms’ working capital from friends and
informal sources. All of these variables introduced together in one regression can lead to
colinearity problems and lower significance of some variables. One possible solution is the
4 We make sure to get a sufficient number of firms for each region in each sector.
10
generation of aggregated indices or choosing just one variable per phenomena. We use Principal
Component Analysis (PCA) and standardization methods to generate the aggregated indices5.
III.2. Results
First, we estimate the impact of an index of physical and financial infrastructure on the
probability for a country to receive FDI (1). We then consider physical infrastructure index
separately from financial infrastructure index (2) and after one variable for each type of
infrastructure (3).
5 Standardization method is similar to PCA but it gives the same weight to all components of the index. Physical
infrastructure index includes firms’ perception of transport and electricity problems as well as telecommunication
opportunities (expressed by firms’ use of e-mail and website in their interaction with their clients and suppliers).
Financial development index includes firms’ perception of access to financing problems and the share of financing
from informal sources (money lender, family and friends) in firms’ working capital.
11
Table 1: Basic model with infrastructure variables
Dependant variable : FDI
(1) (2) (3)
2SLS IV FE Logit 2SLS IV FE Logit 2SLS IV FE Logit
Age -0.002 -0.002 [0.98] -0.002 -0.002 [0.98] -0.002 -0.002 [0.98] (8.09)*** (12.83)*** (8.33)*** (13.55)*** (7.42)*** (11.04)***
Size (20-99 employees) 0.043 0.047 [1.63] 0.044 0.049 [1.66] 0.028 0.033 [1.43] (5.53)*** (9.16)*** (5.11)*** (9.75)*** (3.30)*** (6.05)***
Size (>=100 employees) 0.153 0.140 [3.33] 0.163 0.156 [3.73] 0.129 0.110 [2.69] (9.57)*** (18.75)*** (9.41)*** (20.87)*** (7.74)*** (12.93)***
Agglomeration 0.000 0.000 [1.00] -0.000 0.000 [1.00] 0.000 0.000 [1.00] (0.34) (1.53) (0.04) (0.31) (0.27) (1.95)*
Infrastructure problems -0.136 -0.123 [0.25] (11.10)*** (18.79)***
Physical Infrast. problems -0.265 -0.235 [0.07] (9.91)*** (16.38)***
Financial Infrast. problems -0.026 -0.026 [0.74] (2.60)*** (3.81)***
Telecom problems -0.246 -0.218 [0.09] (9.68)*** (18.37)***
Informal finance -0.001 -0.001 [0.98] (3.15)*** (3.85)***
Observations 33604 33604 33604 33604 33604 33604 Number of countries 77 77 77 77 77 77 R²/Pseudo R² 0.05 0.10 0.05 0.11 0.05 0.10 % of correct prediction 70.63 70.71 70.62
Weak instrument diagnostic Infrastructure/Physical Infrast. Partial R² 0.10 0.03 0.11 Shea partial R² 0.10 0.03 0.11 Partial F 3247.4 190.4 2822.65 [p-value] [0.000] [0.000] [0.000]
Financial Infrast. Partial R² 0.08 0.08 Shea partial R² 0.08 0.08 Partial F 1672.8 61309.5 [p-value] [0.000] [0.000]
Cragg-Donald Stat. 1972.6 374.0 987.6 Critical value (10%) 19.93 13.43 13.43 Clustered z statistics (absolute value) at country level in parentheses All regressions include country and sector fixed effects. The reference for size dummies is small size (less than 20 employees) For logit regression, bootstrapped (with 100 replications) z statistics clustered at country level in parentheses Coefficients reported for logit regression are marginal effects. Next to marginal effects, odds ratios are reported in brackets. * significant at 10%; ** significant at 5%; *** significant at 1% First-stage regressions (not reported for conciseness) are available upon request.
The instrumental logit fixed effect (IV FE logit) estimations imply two stages procedures
leading to consistent parameters but incorrect estimated variances. Through resampling based on
the sample data, we approximate the standard errors by the bootstrap method to obtain proper
standard errors which are reported. Attention is also given to linear probability models which
12
despite some disadvantages approximate well logit specification and allow more flexible handling
of the unobservable heterogeneity and weak instruments diagnostic. Since the explanatory
variables include firms’ perceptions or based on firms’ answers, endogenous issues are serious as
discussed above. The validity of the results depends on the quality of instruments used to address
endogeneity. As instruments diagnostic tests, we rely on some statistics of the first stage estimates
(partial R², Shea partial R², partial F statistic and Cragg-Donald weak instrument test). The
correlations between endogenous variables and excluded instruments are confirmed by the partial
R², which are above zero. Since we have more than one endogenous variable in some cases,
comparison between the standard partial R² and Shea partial R² which take into account the
correlation between instruments can be relevant. These two measures are similar in our case,
indicating low correlation between instruments and therefore no concern as indicated by Baum,
Schaffer and Stillman (2003). In addition, we obtain large and significant F statistics in the first
stage regressions6. A growing weak instrument test in the literature is the comparison of Cragg-
Donald statistics to critical values computed by Stock and Yogo (2004). Our computed Cragg-
Donald statistics are far higher than the Sotck and Yogo critical values, indicating the absence of
the weak instruments problem.
Including only firm level determinants (age and size) and agglomeration effect variable,
basic regressions linking infrastructure problems variables (aggregated indices as well as individual
variables) to FDI show a strong negative relation. These regressions include only the first interest
variables but the theoretical background underlined the importance of other structural factors
such as institutional quality and skilled worker availability. The following estimations take these
additional factors into account and then assess the validity of the preceding results.
6 These statistics are more relevant for the two first estimations in which only one variable is endogenous.
13
Table 2: Basic model including other structural factors
Dependant variable : FDI
(1) (2) (3)
2SLS IV FE Logit 2SLS IV FE Logit 2SLS IV FE Logit
Age -0.002 -0.002 [0.98] -0.002 -0.002 [0.98] -0.002 -0.001 [0.98] (7.92)*** (13.00)*** (8.03)*** (12.5)*** (7.48)*** (11.33)***
Size (20-99 employees) 0.040 0.044 [1.59] 0.038 0.043 [1.57] 0.028 0.033 [1.42] (4.91)*** (7.89)*** (4.44)*** (7.64)*** (3.25)*** (5.90)***
Size (>=100 employees) 0.147 0.133 [3.18] 0.150 0.138 [3.31] 0.128 0.109 [2.69] (9.08)*** (14.64)*** (8.95)*** (17.28)*** (7.78)*** (12.96)***
Agglomeration 0.000 0.000 [1.00] -0.000 -0.000 [1.00] 0.000 0.000 [1.00] (0.31) (1.04) (0.10) (0.83) (0.33) (2.05)**
Infrastructure problems -0.138 -0.125 [0.25] (11.15)*** (17.51)***
Physical Infrast. problems -0.254 -0.225 [0.08] (10.13)*** (19.52)***
Financial Infrast. problems
-0.038 -0.036 [0.67]
(3.88)*** (5.65)***
Telecom problems -0.251 -0.223 [0.08] (9.62)*** (15.80)***
Informal finance -0.001 -0.001 [0.99] (3.21)*** (3.30)***
Skilled labor problems 0.019 0.013 [1.16] 0.043 0.036 [1.49] -0.002 -0.006 [0.94] (1.76)* (2.64)*** (3.30)*** (6.37)*** (0.23) (1.11)
Crime and disorder 0.000 0.001 [1.01] 0.031 0.027 [1.36] -0.022 -0.019 [0.80] (0.05) (0.11) (2.88)*** (4.88)*** (2.48)** (3.96)***
Observations 33604 33604 33604 33604 33604 33604 Number of countries 77 77 77 77 77 77 R²/Pseudo R² 0.05 0.10 0.05 0.11 0.05 0.11 % of correct prediction 70.60 70.78 70.75
Weak instruments diagnostic†
Infrastructure/Physical Infrast. Partial R² 0.11 0.05 0.12 Shea partial R² 0.11 0.04 0.11 Partial F 1610 163 1687 [p-value] [0.000] [0.000] [0.000]
Financial Infrast. Partial R² 0.08 0.08 Shea partial R² 0.08 0.08 Partial F 1099 38206 [p-value] [0.000] [0.000]
Cragg-Donald Stat. 790.9 271.8 590.1 Clustered z statistics (absolute value) at country level in parentheses. All regressions include country and sector fixed effects. The reference for size dummies is small size (less than 20 employees). For logit regression, bootstrapped (with 100 replications) z statistics clustered at country level in parentheses. Coefficients reported for logit regression are marginal effects. Next to marginal effects, odds ratios are reported in brackets. * significant at 10%; ** significant at 5%; *** significant at 1%. † The weak instruments diagnostic tests of other variables (skilled worker problems and crime or disorder) not reported here give partial and Shea partial R² between 0.08 and 0.10 and large F-statistics (far above 10) even if the latter statistics in our cases may not be very relevant for weak instrument diagnostic. Stock and Yogo critical values are available for up to three endogenous regressors. First-stage regressions (not reported for conciseness) are available upon request.
14
Taking into account institutional problems and human capital constraints lead to similar
results for the variables of interest (physical and financial infrastructure). We consider firms’
perception of crime, theft and disorder as a variable for institutional quality and for human capital
variable, we retain firms’ perception of skills and education of available workers. The endogeneity
problem explained above for the interest variables also hold here. Hence, each variable is
instrumented by its sector-region average.
The results show that on average, larger and younger firms are more likely to be foreign.
The agglomeration effect which captures the positive or negative externalities of the number of
foreign firms in one area in the same sector has no effect. Focusing on the interest variables, we
find that the index of physical and financial infrastructure problems negatively and significantly
affects FDI. These results hold using one index by type of infrastructure (physical and financial)
or individual variables. The results are also similar in terms of sign and significance for the two
estimation methods (instrumental variables with fixed-effect logit and simple instrumental
variables approach). Regions with better access to telecommunications, or better access to formal
credit (through the banking system) attract more foreign investments. Indeed, additional
infrastructure problem reduces the odds of receiving FDI by 75%. A breakdown of the general
infrastructure index shows that an additional physical infrastructure problem decreases the
chance of receiving FDI by 92% while the same marginal increase in financial infrastructure
problems reduces the chance of attracting FDI by 33%. Similar conclusions are found with
individual variables for both types of infrastructure. Availability of good roads and transport
facilities, secure provision of electricity and a well-functioning telecommunications system thus
allow and encourage economic activities, especially industrial activities, attracting foreign firms.
Financing opportunities for firms and consumers in local credit markets also encourage foreign
firms to take advantage of these credit possibilities when setting up, operating, or expanding their
business.
15
The additional structural factors (availability of skilled workers and institutional problems)
seem to have a role in FDI attractiveness. Problems regarding the availability of skilled workers
have a positive effect on FDI using aggregated infrastructure indices. Even if this finding is not
robust to different specifications (with non aggregated infrastructure variables), the vertical FDI
theory which states that firms look for cheap low-skilled workers in developing countries support
this result. The effect of institutional problems (crime, theft and disorder) is unstable across the
specifications. We will be back to these two variables (human capital and institution) in the
robustness check.
III.3. Deeper Analyses and Robustness Checks7
FDI theory defines two major types of FDI (vertical and horizontal) with different
motives and therefore different determinants. Using the share of exports in firms’ sales, a
breakdown analysis between exporting and non-exporting firms helps to get a deeper
understanding of the importance of infrastructure for each type of FDI. According to the
structure of ownership, foreign firms’ criteria to invest in a country could differ. Foreign firms
may prefer a joint venture with local partners in order to reduce risk when investing in a foreign
country. Even with the inclusion of sector fixed effects in all regressions, a breakdown analysis by
sector gives a more complete picture of the importance of infrastructure for manufacturing
sector compared to services or the effect of infrastructure across the heterogeneous
manufacturing sector. Another robustness check assesses the validity of the results after inclusion
of additional explanatory variables reflecting institutions, taxes, regulation policies and market
size. The last robustness check compares the poorest group of countries (Sub Saharan Africa) to
the rest of the developing countries and analyzes heterogeneity in the impact of structural
variables across countries.
7 Only 2SLS regressions results which are similar (sign and significance) to those obtained with logit specification are
reported for these deeper analyses and robustness checks.
16
Breakdown by export status: Exporter versus non exporter firms
This breakdown allows testing for the difference between local market-oriented FDI
(horizontal FDI) and export-oriented FDI (vertical FDI) and assessing how investment climate
affects horizontal FDI versus vertical FDI8. As suggested by the theory, horizontal FDI is more
affected by market potential and vertical FDI by factor costs (mainly unit labor cost or wage).
Considering the breakdown by FDI type, the results show that physical infrastructure and
financial development are important, both for the attractiveness of exporter foreign firms or
foreign firms which sale their production locally. Foreign firms exporting their production are
more affected by telecommunication problems compared to firms supplying the local market
(appendix 3). These results indicate that infrastructure problems are serious obstacles. For
example, telecommunication problems may impede interactions between suppliers and their
customers. Lack of financial development affects firms and the economy in general as explained
in the theoretical section. Foreign firms selling their productions locally are therefore more
affected by financing constraints compared to firms exporting their production as shown by the
results. Indeed, exporter and non-exporter foreign firms support the financing constraints during
the production process but non-exporter firms also support the financing constraints of the
economy in general which affect their clients. Exporter firms are also more affected by skilled
worker shortage compared to firms supplying the local market when institutional problems
(disorder, crime and theft) affect both groups of firms. These results are similar with exporter
firms defined as firms exporting at least 10% of their production or as firms exporting any part of
their production.
8 This paper has the strength to address this issue even if some aspects of the breakdown (direction of affiliate sales
which are not available in our data) should also be considered.
17
Analysis across ownership degree: from local to joint-venture and foreign fully-owned firms
Foreign firms’ location factors may vary with ownership degree. All of the analyses above
were based on a dichotomous variable which takes into account only the fact that the firm is local
or foreign regardless of the degree of foreign ownership which can range from fully local firm to
joint venture and fully foreign firm. This section takes this variability into account and assesses
how investment climate, especially physical and financial infrastructure can affect the degree of
foreign ownership. The hypothesis is that the impact of infrastructure varies and is greater with
ownership degree. Foreign firms may look for a local partner in a joint venture when they plan to
invest in a country with important infrastructure constraints or political instability. Firms’
investments through a joint venture partnership aim to reduce information costs of local market
(in the special case of new firms). To analyze this question, we use 2SLS and two limits tobit with
instrumentation9.
The results show that physical and financial infrastructure problems negatively and
significantly affect foreign ownership. In regions with infrastructure and financing problems,
foreign investors participate less in firms’ capital. The results also highlight that institutional
problems (crime, theft and disorder) reduce the share of foreign participation in firms’ capital
(appendix 3).
Breakdown by sector
Factors influencing firms’ location may vary across sectors. Firms in service sector could
be more attracted by human capital (high skilled workers) availability when manufacturing firms
could be more attracted by good infrastructure provision. The manufacturing sector is also
heterogeneous so that different industries in this aggregated sector could be differently affected
9 This method is more relevant because of the great number of firms which are fully owned by local investors or by
foreign investors leading to an important number of observations in upper and lower limit of the distribution of
foreign ownership variable. Standard OLS would lead to downward bias in the predicted ownership degree.
18
by the same factors (especially physical and financial infrastructure). All the regressions included
sector fixed-effects in order to take into account this heterogeneity. This section helps to check if
the aggregated (all sectors combined) impact of infrastructure is driven by one or two sectors, or
if all of the sectors show the same pattern in the importance of infrastructure for foreign firms
attractiveness (appendix 4). In most of the sectors, availability of well functioning infrastructure
increases the probability of receiving FDI. Physical and financial infrastructure problems decrease
the probability of receiving FDI in four sectors: textile, garment, metal and machinery and retails.
Foreign firms’ location in agro-industry, electronics, wood & furniture, non metallic & plastic
materials and services sectors are negatively and significantly affected by physical infrastructure
when financial development affects firm location in the leather sector. Only FDI in chemical
sector is not affected by physical infrastructure and financial development. These results show
that infrastructure problems are important constraints regardless of the sectors, and they
highlight the crucial role of physical infrastructure for FDI attractiveness in developing countries.
Additional control variables
As mentioned in previous section of the paper, explanatory variables have been chosen
according to their economic relevance and also considering their number of non missing values.
This robustness check includes a greater number of control variables leading to more missing
values. Firms’ perception of crime, theft and disorder in the preceding regressions (baseline
regression including more structural factors) is replaced by a variable closer to the quality of
institution in line with economic activity: the property right protection10. The additional control
variables are firms’ perceptions of some investment climate constraints. These variables are
consecutively: labor regulation, corruption, customs and trade regulations, tax rates and wages
10 Introduction of property right protection variable leads to bigger missing values compared to the variable of theft,
disorder and crime. This is not a concern in this section since the additional control variables lead to approximately
the same number of missing values.
19
(appendix 5). The last variable (wages) takes into account the labor cost and is measured by
sector-region averages of wage per employee. The endogeneity problems explained above still
hold for the new control variables. Each variable is then instrumented by its sector-region
average. Additional explanatory variables help also to overcome endogeneity problems due to
omitted variables bias. Including the additional variables cumulatively, physical and financial
infrastructure problems remain significant and negatively affect FDI attractiveness. When
included, corruption and tax rate problems negatively and significantly affect FDI. However
customs and trade regulation problems increase FDI attractiveness. This result is supported by
the horizontal FDI theory. Since horizontal FDI aims to supply the local market, the theory
suggests that trade barriers may be indirect protections for firms located in the country giving
them price advantages.
Sub-Saharan Africa specificity
As mentioned in the literature review, most analyses have focused on either an Asian or
Latin American developing country. The enterprise surveys allow us to perform an analysis
including an important number of African countries, the least developed and the ones with the
highest investment climate constraints. This study thus gives the first picture of FDI location in
Africa using firm level data. Physical infrastructure problems negatively and significantly affect
FDI in Sub-Saharan African (SSA) countries and in other developing countries (appendix 6). The
size of this negative impact is very close across the two groups (SSA countries and other
developing countries). Availability of well-functioning telecommunications systems increases the
probability of receiving FDI in SSA and other developing countries. However, lack of financial
development discourages FDI in SSA countries and other developing countries but the size of
the negative effect is bigger for SSA. Thus, an improvement in investment climate with better
access to finance increases more FDI in SSA relatively to other developing countries. Social
instability captured by theft, disorder and crime problems are more relevant for SSA countries in
20
FDI attractiveness compared to other developing countries even if this factor discourages FDI in
all regions. These results are consistent with expectations, since compared to other developing
countries, many SSA countries are engaged in civil war and violent protests leading sometimes to
crimes and murders. As indicated in previous sections, trade and customs regulations encourage
FDI following horizontal FDI theory. Except in SSA countries, tax rate is an investment climate
constraint that discourages FDI in other developing countries. This finding supports the
theoretical hypothesis according to which tax incentives should have lower impact in developing
countries (compared to developed countries), given their structural problems (infrastructure,
institution, etc.). Since SSA is the less developed of developing countries, tax incentives then have
lower effect in this region compared to other developing countries.
Appendix 7 analyzes the heterogeneity in the impact of major structural factors on FDI
attractiveness at country level. In almost all of the countries, investment climate constraints
(especially physical and financial infrastructure problems) reduce the probability to host foreign
firms. These findings highlight the fact that the results are not driven by particular country or
group of countries.
Conclusions
Foreign direct investment attractiveness is a key issue in economics and business in the
developing and developed world. This paper analyzes how investment climate constraints
jeopardize FDI attractiveness in developing countries. Using firm level data for 77 developing
countries, this paper provides the first empirical analysis of the importance of investment climate
in FDI attractiveness with a large sample of developing countries. For investment climate
constraints, we focused on physical and financial infrastructure problems in addition to human
capital and institutional constraints. The main results show that improving physical and financial
infrastructure increases the probability of receiving a foreign firm (FDI).
21
A breakdown analysis between exporter and non-exporter firms shows that firms
supplying foreign markets are more affected by physical infrastructure problems but financing
constraints affect more foreign firms supplying local markets. Infrastructure problems remain a
key constraint with alternative definition of foreign ownership, analysis by sector and after
inclusion of additional explanatory variables. Physical infrastructure problems are revealed to be a
major constraint for foreign firms in SSA and other developing countries when financing
constraints are more serious obstacles for foreign firms in African countries. The results also
highlight the importance of institutional quality in attracting FDI to developing countries
especially to Sub Saharan African countries.
When designing their policies to attract foreign investment, developing countries should
pay particular attention to infrastructure (physical and financial) and institutions given the major
role of these factors in FDI attractiveness. These policies should also take into account the
potential negative impacts of FDI on local firms and local economy.
22
Bibliography
Baum C. F., Schaffer M. E. and Stillman S., 2003, “Instrumental variables and GMM:
Estimation and testing”, Stata Journal, 3(1): 1-31.
Bloningen B. A., 1997, “Firm Specific Assets and the Link Between Exchange Rates and
Foreign Direct Investment” American Economic Review, 87(3): 447-65
Bloningen B. A. and Ronald B. D., 2004, “The Effects of Bilateral Tax Treaties on US FDI
Activity”, International Tax and Public Finance, 11(5): 601-22.
Carr D., Markusen J. R. and Maskus K., 2001, “Estimating the Knowledge Capital Model
of the Multinational Firm”, American Economic Review, 91: 693-708.
Cheng L. K. and Kwan Y. K., 2000, “What are the determinants of the location of foreign
direct investment? The Chinese experience”, Journal of International Economics, 51: 379–400.
De Mooij R. A. and Ederveen S., 2003, “Taxation and Foreign Direct Investment: A
Synthesis of Empirical Research”, International Tax and Public Finance, 10(6): 673-93.
Deichmann J., Karidis S. and Sayek S., 2003, “Foreign direct investment in Turkey:
Regional determinants”, Applied Economics, 35(16):1767-1778.
Desai M. A., Fritz C. F. and James R. H., 2004, “Foreign Direct Investment in a World of
Multiple Taxes”, Journal of Public Economics, 88(12): 2727-44.
Disdier A. C. and Mayer T., 2004, “How Different in Eastern Europe? Structure and
Determinants of Location Choices by French Firms in Eastern and Western Europe”, Journal of
Comparative Economics, 32: 280-296.
Du J., Lu Y. and Tao Z., 2007, “Economic Institutions and FDI Location Choice:
Evidence from U.S. Multinationals in China” Mimeo.
Dunning J. H., 1980, “Toward an eclectic theory of international production: some
empirical tests’”, Journal of International Business Studies, 11 (1): 9-31.
23
Dunning J. H., 1993, “Multinational Enterprises and the Global Economy”, Addison-
Wesley Publishing Company: Reading, MA.
Dunning J. H., 1998, “Location and the multinational enterprise: a neglected factor?”
Journal of International Business Studies, 29(1): 45-66.
Fung K. C., Iizaka H. and Parker S., 2002, “Determinants of U.S. and Japanese foreign
direct investment in China”, Journal of Comparative Economics, 30(3): 567–578.
Gao T., 2005, “Labor quality and the location of foreign direct investment: Evidence from
China”, China Economic Review, 16(3): 274-292.
Hahn J., and Hausman J. A., 2002, “A New Specification Test for the Validity of Instrumental
Variables”, Econometrica, 70(1): 163-189.
Hanson , G., Mataloni, R. J., and Slaughter M., 2001, “Expansion Strategies of U.S.
Multinationals Firms”. In Brookings Trade Forum 2001, ed. D. Rodrik and S. Collins, pp. 245-
282.
Jenkins C. and Thomas L., 2002, “Foreign Direct Investment in Southern Africa:
Determinants, Characteristics and Implications for Economic Growth and Poverty Alleviation”,
Mimeo.
Julan Du, Yi Lu and Zhigang Tao, 2007, “Economic Institutions and FDI Location
Choice: Evidence from U.S. Multinationals in China” Mimeo.
Kang S., Lee H., 2004, “Location Choice of Multinational Companies in China: Korean
and Japanese Companies” Finance Working Papers (East Asian Bureau of Economic Research), No
529.
Navaretti G. B. and Venables A. J., 2006, “Multinational Firms in the World Economy”
Princeton University Press.
24
Norbäck P. J., 2001, “Multinational Firms, Technology and Location”, Journal of International
Economics, 54: 449-469.
Shea J., 1997, “Instrument Relevance in Multivariate Linear Models: A Simple Measure”,
Review of Economics and Statistics 79: 348-352.
Stock J. H. and Yogo M., 2004, “Testing for Weak Instruments in Linear IV Regressions”,
Mimeo.
Sun Q., Tong W. and Yu Q., 2002, “Determinants of Foreign Direct Investments across
China”, Journal of International Money and Finance, 21(1): 79-113.
Urata S. and Kawai H., 2000, “The Determinants of the Location of Foreign Direct
Investment by Japanese Small and Medium-sized Entreprises”, Small Business Economics 15 79-103.
Yeaple S., 2003, “The role of skill endowments in the structure of US outward foreign
investments”, Review of Economics and Statistics, 85: 726-734.
World Bank, 2006, “Governance, Investment Climate, and Harmonious Society:
Competitiveness Enhancements for 120 Cities in China”. Report No. 37759-CN.
25
Appendices
Appendix 1: List of variables
Variable Definitions
FDI Dummy equal 1 if at least 10% of firm capital is foreign Age Firm age Size Firm Size: 3 categories based on permanent & temporary workers Agglomeration Number of foreign firms in same sector and same region Telecom problems 1 Use of e-mail for business with clients & suppliers (dummy variable) Telecom problems 2 Use of website for business with clients & suppliers (dummy variable) Electricity problems Business constraint: electricity Transport problems Business constraint: transport Informal finance problems Informal sources of financing in firms’ working capital Access to finance problems Business constraint: access to finance (e.g. collateral) External auditor Annual financial statement reviewed by external auditor (dummy variable) Skilled labor problems Business constraint: skills of available workers Crime and disorder Business constraint: crime, theft, disorder Property right Confident judicial system will uphold property rights Labor regulation Business constraint: labor regulations Corruption Business constraint: corruption Custom and trade Business constraint: customs and trade regulations Tax rates Business constraint: tax rates Wage Sector-region average wage per employee
Appendix 2: Descriptive statistics
Variable Mean Std. Dev. Minimum Maximum Observation
FDI 0,14 0,35 0 1 33604
Telecom problems 1 1,36 0,48 1 2 33604
Telecom problems 2 1,55 0,50 1 2 33604
Electricity problems 0,87 1,24 0 4 33604
Transport problems 0,62 1,05 0 4 33604
Informal finance problems 5,03 16,72 0 100 33604
Access to finance problems 1,13 1,34 0 4 33604
External auditor 1,49 0,50 1 2 33085
Skilled labor problems 0,92 1,18 0 4 33604
Property right 3,55 1,45 1 6 23500
Labor regulation 0,82 1,15 0 4 32226
Corruption 1,18 1,42 0 4 33145
Custom and trade 0,85 1,22 0 4 31698
Tax rates 1,40 1,39 0 4 33491
Age 17,08 17,50 0 202 33604
Agglomeration 27,04 28,36 0 144 33604
Wage 179,65 4554,57 0 400691 15685
Number of permanent workers 138,26 495,48 0 19047 33471
26
Appendix 3: Breakdown by export status and foreign ownership
Dependant variable: FDI
Dependant variable: foreign ownership
2SLS 2SLS 2SLS 2SLS 2SLS IV Tobit
(1) (2) (3)
Non-exporter* Exporter Non-exporter Exporter
Telecom problems -0.191 -0.281 -0.187 -0.280 -0.204 -2.843 (7.26)*** (4.59)*** (7.25)*** (4.86)*** (8.51)*** (17.70)***
Informal finance -0.001 -0.003 -0.001 -0.003 -0.001 -0.017 (1.73)* (1.24) (1.58) (1.48) (2.71)*** (3.67)***
Skilled labor problems 0.004 -0.048 0.006 -0.037 -0.003 -0.083 (0.46) (2.50)** (0.78) (1.95)* (0.33) (1.18)
Crime and disorder -0.020 -0.030 -0.018 -0.041 -0.017 -0.222 (2.86)*** (1.27) (2.47)** (2.09)** (2.29)** (3.28)***
Age -0.001 -0.004 -0.001 -0.003 -0.002 -0.018 (5.05)*** (7.87)*** (5.89)*** (6.80)*** (7.63)*** (13.66)***
Size (20-99 employees) 0.020 0.084 0.019 0.070 0.018 0.392 (2.71)*** (4.45)*** (2.57)** (4.34)*** (2.41)** (6.33)***
Size (>=100 employees) 0.078 0.199 0.068 0.178 0.100 1.171 (4.87)*** (9.19)*** (4.14)*** (9.27)*** (6.68)*** (13.51)***
Agglomeration 0.000 -0.000 0.000 -0.000 0.000 0.003 (1.59) (0.50) (1.37) (0.22) (0.39) (3.09)***
Observations 26460 7031 24543 9061 33604 33604 Number of countries 77 77 77 77 77 77
Weak instrument diagnostic+ Telecom problems Partial R² Shea partial R² Partial F p-value
0.11 0.10 1039 [0.000]
0.08 0.08 36 [0.000]
0.10 0.10 884 [0.000]
0.08 0.08 39 [0.000]
0.12 0.11 1687 [0.000]
0.12 0.11 1687 [0.000]
Informal finance Partial R² Shea partial R² Partial F p-value
0.08 0.08 912 [0.000]
0.06 0.06 27 [0.000]
0.08 0.08 771 [0.000]
0.07 0.07 35 [0.000]
0.08 0.08 38206 [0.000]
0.08 0.08 38206 [0.000]
Cragg-Donald Stat. 475.9 77.9 441.6 122.7 590.1 Clustered z statistics (absolute value) at country level in parentheses All regressions include country and sector fixed effects For tobit regression, bootstrapped (with 100 replications) z statistics clustered at country level in parentheses. The tobit regression include countries and sector dummies The reference for size dummies is small size (less than 20 employees) * significant at 10%; ** significant at 5%; *** significant at 1% * Exporter are defined as firms exporting at least 10% of their sales in (1) and as firms exporting any part of their sales in (2). + The weak instruments diagnostic tests of other control variables (additional structural factor: skilled worker problems and crime or disorder) not reported here give partial and Shea partial R² between 0.08 and 0.10 and large F-statistics. First-stage regressions (not reported for conciseness) are available upon request.
Appendix 4: Estim
ations by sector
Dependant variable : FDI
Textile
Leather
Garm
ent
Agro
Metal & M
ac
Electronics
Chemical
Wood
Non-m
etal.
Retails
Services
Telecom problems
-0.284
0.097
-0.211
-0.241
-0.224
-0.711
-0.066
-0.132
-0.212
-0.245
-0.363
(3.82)***
(1.02)
(2.64)***
(5.92)***
(3.88)***
(2.64)***
(0.78)
(2.21)**
(3.24)***
(6.10)***
(7.11)***
Informal finance
-0.004
0.004
-0.004
0.000
-0.003
-0.005
0.003
0.001
-0.002
-0.002
0.001
(3.92)***
(2.58)***
(5.09)***
(0.14)
(2.17)**
(0.99)
(1.29)
(1.19)
(0.86)
(2.27)**
(0.48)
Skilled labor prob.
-0.047
-0.014
-0.009
-0.015
0.026
0.026
0.069
-0.011
-0.016
0.018
-0.039
(1.20)
(0.41)
(0.34)
(0.73)
(0.83)
(0.19)
(1.72)*
(0.43)
(0.53)
(0.59)
(1.30)
Crime and disorder
0.039
-0.045
-0.067
0.004
-0.025
0.040
-0.019
-0.020
-0.055
-0.002
-0.039
(1.63)
(1.71)*
(1.98)**
(0.29)
(0.85)
(0.65)
(0.41)
(1.32)
(1.56)
(0.08)
(1.38)
Age
-0.002
-0.001
-0.002
-0.001
-0.002
-0.002
0.000
-0.001
0.000
-0.002
-0.002
(2.23)**
(0.92)
(3.55)***
(3.19)***
(5.41)***
(1.17)
(0.38)
(2.37)**
(0.34)
(10.13)***
(5.52)***
Size (20-99 empl.)
0.025
0.089
0.010
0.022
0.046
-0.033
0.055
0.061
0.003
0.034
0.032
(1.29)
(2.84)***
(0.46)
(1.21)
(2.54)**
(0.38)
(2.12)**
(2.60)***
(0.10)
(2.43)**
(1.83)*
Size (>=100 empl.)
0.078
0.125
0.164
0.110
0.187
0.222
0.239
0.147
0.062
0.059
0.059
(1.57)
(2.27)**
(3.89)***
(3.26)***
(6.65)***
(1.62)
(4.87)***
(3.62)***
(1.25)
(2.52)**
(1.87)*
Agglomeration
-0.001
0.004
-0.000
0.000
-0.001
-0.002
0.000
0.000
-0.001
0.001
0.001
(1.43)
(2.69)***
(0.30)
(1.84)*
(1.24)
(3.48)***
(0.19)
(0.15)
(2.49)**
(2.84)***
(1.72)*
Weak instrument
diagnostic+
Cragg-Donald Stat. 16.3
3.0
24.0
74.7
38.3
1.5
15.3
37.3
11.5
44.7
50.1
Observations
1928
494
3702
5638
3642
500
1661
2870
1474
5702
5988
Number of countries
52
28
62
76
69
9 53
67
54
49
50
Clustered z statistics (absolute value) at country level in parentheses
All regressions include country and sector fixed effects
The reference for size dummies is small size (less than 20 employees)
* significant at 10%; ** significant at 5%; *** significant at 1%
+ The weak instruments diagnostic tests give partial and Shea partial R² between 0.04 and 0.10 and large F-statistics.
First-stage regressions (not reported for conciseness) are available upon request.
Appendix 5: Estimation with additional variables
Dependant variable : FDI
(1) (2) (3) (4) (5) (6) (7)
Telecom problems -0.251 -0.278 -0.281 -0.284 -0.273 -0.270 -0.231 (9.62)*** (9.80)*** (10.19)*** (10.22)*** (9.60)*** (9.62)*** (6.59)***
Informal finance -0.001 -0.002 -0.002 -0.001 -0.002 -0.002 -0.002 (3.21)*** (3.03)*** (2.96)*** (2.81)*** (3.08)*** (2.85)*** (4.98)***
Skilled labor problems -0.002 -0.018 -0.021 -0.017 -0.026 -0.025 -0.024 (0.23) (1.69)* (1.94)* (1.56) (2.25)** (2.16)** (1.81)*
Crime and disorder -0.022 (2.48)**
Property right 0.005 0.005 0.004 0.006 0.005 0.009 (0.64) (0.66) (0.59) (0.76) (0.70) (0.85)
Age -0.002 -0.002 -0.002 -0.002 -0.002 -0.002 -0.001 (7.48)*** (6.45)*** (6.83)*** (6.84)*** (6.91)*** (6.85)*** (4.24)***
Size (20-99 empl.) 0.028 0.027 0.026 0.027 0.025 0.023 0.019 (3.25)*** (2.72)*** (2.57)** (2.66)*** (2.70)*** (2.44)** (1.71)*
Size (>=100 empl.) 0.128 0.119 0.120 0.119 0.113 0.108 0.130 (7.78)*** (6.42)*** (6.64)*** (6.84)*** (6.48)*** (6.37)*** (6.02)***
Agglomeration 0.000 -0.000 -0.000 -0.000 -0.000 -0.000 -0.000 (0.33) (0.20) (0.24) (0.22) (0.19) (0.19) (0.80)
Labor regulation 0.005 0.012 0.003 0.006 0.010 (0.44) (0.90) (0.20) (0.41) (0.54)
Corruption -0.021 -0.031 -0.027 -0.031 (2.24)** (3.28)*** (2.76)*** (3.21)***
Custom and trade 0.042 0.048 0.035 (4.72)*** (4.95)*** (3.28)***
Tax rates -0.021 -0.015 (2.19)** (1.34)
Wage -0.002 (0.23)
Weak instrument diagnostic+
Cragg-Donald Stat. 590.1 426.0 329.1 273.9 224.4 198.2 10.8
Observations 33604 23500 23130 22749 21170 21105 13576 Number of countries 77 55 55 55 55 55 51 Clustered z statistics (absolute value) at country level in parentheses All regressions include country and sector fixed effects The reference for size dummies is small size (less than 20 employees) * significant at 10%; ** significant at 5%; *** significant at 1% +The weak instruments diagnostic tests give partial and Shea partial R² between 0.08 and 0.12 and large F-statistics. First-stage regressions (not reported for conciseness) are available upon request.
29
Appendix 6: African specificity Dependant variable : FDI
Sub-Saharan Africa Other developing countries
(1) (2) (3) (4) (5) (6)
Telecom problems -0.238 -0.239 -0.224 -0.248 -0.248 -0.229 (3.22)*** (3.26)*** (2.53)** (8.54)*** (8.63)*** (7.86)***
Informal finance -0.000 -0.001 (0.51) (3.09)***
Skilled labor problems -0.019 -0.013 -0.014 0.001 0.006 -0.009 (0.73) (0.54) (0.60) (0.10) (0.53) (0.69)
Crime and disorder -0.038 -0.034 -0.049 -0.020 -0.014 -0.018 (1.71)* (1.62) (2.57)** (2.11)** (1.26) (1.34)
Age -0.001 -0.001 -0.001 -0.002 -0.002 -0.002 (1.00) (1.30) (1.14) (7.78)*** (7.82)*** (8.11)***
Size (20-99 employees) 0.065 0.064 0.058 0.022 0.023 0.022 (2.12)** (2.12)** (1.55) (2.57)** (2.71)*** (2.58)***
Size (>=100 employees) 0.221 0.203 0.194 0.118 0.118 0.110 (5.31)*** (4.93)*** (3.96)*** (6.58)*** (6.90)*** (6.68)***
Agglomeration -0.000 -0.000 -0.000 0.000 0.000 0.000 (0.57) (0.61) (0.51) (0.40) (0.37) (0.33)
Access to finance probl. -0.042 -0.053 -0.027 -0.032 (2.00)** (2.23)** (2.68)*** (3.03)***
Labor regulation -0.039 0.014 (1.02) (0.92)
Corruption 0.028 -0.012 (1.35) (1.04)
Custom and trade 0.042 0.058 (2.62)*** (5.22)***
Tax rates 0.001 -0.020 (0.08) (1.87)*
Weak instrument diagnostic+
Cragg-Donald Stat. 73.4 75.3 28.5 497.0 489.6 220.2
Observations 5366 5366 4150 28238 28238 25861 Number of countries 23 23 22 54 54 54 Clustered z statistics (absolute value) at country level in parentheses All regressions include country and sector fixed effects The reference for size dummies is small size (less than 20 employees) * significant at 10%; ** significant at 5%; *** significant at 1% +The weak instruments diagnostic tests give partial and Shea partial R² between 0.07 and 0.12 and large F-statistics. First-stage regressions (not reported for conciseness) are available upon request.
30
Appendix 7: Heterogeneity in the impact of major structural problems Telecom problems and FDI location Informal finance and FDI Location
ALB
ARM AZE
BGR
BIH
BLR
BRA
CHL
CRI
CZE
ECU
GEO
HND
HRVHUN
IDN
KAZ
KEN
KGZ
KHM
LKA
MDA
MDG
MKD
MUS
MWI
NICPAK
PHL
POL
ROM
RUS
SLVTHA
TJK
TUR
UGAUKR
UZB
VNM
YUGZAF
ZMB
02
46
81
0A
bsol
ute
t-s
tatis
tic
-4 -2 0 2 4Coefficient
ARM
AZE
BEN
BGD
BGR
BIH
BLRBRA
CHLCRI
CZEECU EST
GEO
GTMGUY
HND
HRV
HUN
IDNKAZ
KEN
KGZ
KHM
LKALVA
MDA
MDG
MUSMWI
NIC
PAKPHL
POL
ROM
RUS
SLV
SVKSVN
THA
TJK TUR UGA
UKR
UZB
VNMYUG ZAF
ZMB
02
46
81
01
2A
bsol
ute
t-s
tatis
tic
-2 0 2Coefficient
Corruption problems and FDI Location
Skilled labor problems and FDI location
ALB
ARM
AZE
BEN
BGD
BGRBIH
BLR
BRA
CHL
CRICZE
ECU
EST
GEO
GTM
GUY
HND
HRV
HUN
IDN
KAZ
KEN
KGZ KHM
LKA
LTU
LVA
MDA
MDGMUS
MWI
NICPAK
PHL
POL
ROM
RUSSLV
SVK
SVN
THA
TJK
TUR
UGA
UKR
UZB
VNM
YUG
ZAF
02
46
8A
bsol
ute
t-s
tatis
tic
-4 -2 0 2 4Coefficient
ARM
AZE
BGD
BGR
BIH BLRBRACHLCRI
CZEECUEST
GEO
GUY
HND
HRV
HUN
IDN
KAZ
KEN
KGZ
KHM
LKA
LVA
MDAMDGMKD
MUSMWI
NIC
PAKPHL
POL
ROM
RUS
SLV
SVK
SVN
THA
TJK
TURUGA
UKR
UZBVNMYUG
ZAF
ZMB
05
10
15
20
Abs
olu
te t
-sta
tistic
-1.5 -1 -.5 0 .5Coefficient
These graphs resume coefficients and t-statistics of country level estimations. Each point represents the effect of the structural factors (telecom problems, informal finance, corruption problems and skilled labor problems) on FDI for one country. Coefficients are reported on the horizontal axis and the absolute value of t-statistics on the vertical axis. All regressions have FDI variable as dependant variable and structural factors as explanatory variables. The regressions also include firm characteristics and other control variables and deal with endogeneity issue as regressions (3) and (6) of appendix 6 (graphics are country-level illustration of regression (3) and (6) of appendix 6). The red line t=1.64 indicates 10% significance level. The bleu line indicates the null value of coefficient. The upper-left side of all graphs indicates countries for which structural factor constraints reduce significantly FDI attractiveness.
31
Appendix 8: List of countries
Asia Eastern and Central Europe Latin America and Carribean Sub-Saharan Africa
Bangladesh Albania Argentina Angola Cambodia Armenia Bolivia Benin Indonesia Azerbaijan Brazil Botswana Lebanon Belarus Chile Burundi Mongolia Bosnia and Herzegovina Colombia Congo. Dem. Rep. Pakistan Bulgaria Costa Rica Eritrea Philippines Croatia Ecuador Ethiopia Sri Lanka Czech Republic El Salvador Gambia Thailand Estonia Guatemala Guinea-Bissau Vietnam Georgia Guyana Kenya Hungary Honduras Lesotho Kazakhstan Mexico Madagascar Kyrgyz Republic Nicaragua Malawi Latvia Panama Mali Lithuania Paraguay Mauritania Macedonia. FYR Peru Mauritius Moldova Uruguay Namibia Poland Senegal Romania South Africa Russian Federation Swaziland Serbia and Montenegro Tanzania Slovak Republic Uganda Slovenia Zambia Tajikistan Turkey Ukraine Uzbekistan